Comparison of External, Internal and Construct Validity

Introduction

 Organizations that make the decision to take on additional moral obligations to maximize the firm’s positive effect on society, while reducing the negative impact, are considered to be organizations concerned with implementing corporate social responsibility (CSR) initiatives (Persons 2012). CSR includes voluntary disclosure reporting and practices that a firm implements to improve the well-being of its society (Mishra & Modi 2016). CSR disclosure reporting can positively impact corporate reputation, enhancing stakeholders, and the operating society’s view of the organization (Axjonow, Ernstberger, & Pott 2018).

 This paper will contrast three types of validity: external validity, internal validity, and construct validity. It will identify the different threats to internal validity, external validity, and construct validity. Lastly, this paper will address the impact either of these forms of validity will have on the envisioned research. 

Types of Validity

 External validity involves the generalizations of findings made by researchers in hope that their research will be relevant for other populations, settings, etc. (Cozby 2014). Internal validity is instead focused on the structure of a study and the accuracy of the conclusions drawn based on a cause and effect relationship (Andrade 2018; Cozby 2014). Construct validity answers questions about the measurement of a concept or construct (Cozby 2014). It focuses on identifying if the selected concepts or constructs are suitable for the selected variables (Cozby 2014).

 External validity and internal validity often conflict (Cozby 2014). Since internal validity is based on a cause and effect relationship between specified variables, if that is what the researcher seeks to identify, then there is a strong case for establishing internal validity (Cozby 2014). Certain forms of internal validity involve using a manipulated dependent variable that can cause another dependent variable (Cozby 2014). The manipulation of those variables leads to the fabrication of an environment, which limits the possibility for generalizations to be made (Cozby 2014).

For example, lab experiments have stronger internal validity than external validity because the environment is manipulated into what the researcher wants it to be. The use of random samples instead of non-random samples produces stronger external validity because it takes away possible bias (Cozby 2014).  When there is good evidence that external validity is strong, it makes generalizability possible (Trochim, Donnelly, & Arora 2016).

External Validity

 Two common approaches to external validity are the sampling model and the proximal similarity model. The sampling model identifies the population that the researcher wants to draw the generalizations from, and selects the sample from that population (Trochim, Donnelly, & Arora 2016). This makes drawing generalizations easier because the sample was selected from the desired population. The proximal similarity model uses the closest and most related settings to form generalizations (Trochim, Donnelly, & Arora 2016).

A researchers confidence in external validity can increase if the findings of the research are applicable to other settings, populations, etc. (Cozby 2014). The findings can be replicated through exact replication or conceptual replication. Exact replication replicates the exact setting, population, measures, etc. that were used for the initial research (Cozby 2014). Conceptual replication replicates the findings by way of different procedures, settings, population (Cozby 2014).

Internal Validity

Strong internal validity stems from the conclusions drawn which imply that one variable ultimately changed another (Cozby 2014). Internal validity is used to prove that there is truth in the assumptions drawn on cause and effect relationships (Trochim, Donnelly, & Arora 2016). Strong internal validity can be formed from the analysis of three elements. The temporal precedence, covariation between two variables, and the need to eliminate plausible alternative explanations for the observed relationship are the three elements used to establish strong internal validity (Trochim, Donnelly, & Arora 2016).

The experimental method is a way to establish strong internal validity. It reduces the ambiguity related to how the results are interpreted with one variable being manipulated (dependent variable), and one independent variable (Cozby 2014).  Like external validity randomization in selecting a sample enhances internal validity. Randomization makes sure that any confounding variable is likely to affect all identified groups (Cozby 2014).

Construct Validity

Construct validity is concerned with the operational definition of variables, with the operational definition reflecting the theoretical meaning of those variables (Cozby 2014). Construct validity wants to identify if the measures being used to measure the identified constructs are right for those constructs (Cozby 2014). It contributes to the quality of the measurement and the overarching category of this kind of validity (Trochim, Donnelly, & Arora 2016).

There are different types of validity that stem from construct validity, which Trochim, Donnelly, & Arora (2016) identified as being grouped either under what they call translation validity or criterion-related validity. Translation validity identifies if the defined measurement factors being used are a good translation for the construct (Trochim, Donnelly, & Arora 2016). Face validity and content validity are branches under translation validity. Face validity suggests that the measures being used appear to accurately assess the intended variables (Cozby 2014). The content validity compares the content of measure with the content that defines the construct (Cozby 2014).

Criterion-related validity focuses on if the measurable factors for a construct behaves how it should (Trochim, Donnelly, & Arora 2016). Predictive validity, concurrent validity, convergent validity, and discriminant validity stem from criterion-related validity. Predictive validity is the use of a measure to predict the future behavior of a construct (Cozby 2014). Concurrent validity is demonstrated through research used to examine the relationship between the measures and a specified behavior concurrently (Cozby 2014). Convergent validity compares the scores of the questioned measure to the scores of other measures with similar constructs (Cozby 2014). Discriminant validity measures do not relate to variables or constructs (Cozby 2014).

Threats

Internal Validity

 Some threats to internal validity are history, maturation, testing, instrumentation, and morality (Trochim, Donnelly, & Arora 2016). The threat of history is similar to the threat of timing for external validity. The threat of history occurs when another event occurs at the same time as the researcher’s research. It threatens the plausibility of cause and effect stating that the other random event could have caused the outcome instead of the identified independent variable (Trochim, Donnelly, & Arora 2016). The threat of maturation occurs when there is an outcome due to something running its natural course, and over time it occurs, instead of the identified variables being the cause of the outcome (Trochim, Donnelly, & Arora 2016).

 Another threat to internal validity is related to experimental and nonexperimental research which creates the direction for the cause and effect relationship (Cozby 2014). With nonexperimental research, it is difficult to determine which variable causes the other and creates the possibility of a third variable (Cozby 2014). Confounding variables, which are the third variables, can be infinite and are difficult to keep constant (Cozby 2014).

External Validity

A threat to external validity exists within the people, place, or time used to conduct the research and form the generalizations (Trochim, Donnelly, & Arora 2016). The generalizations formed could be skewed due to the group of people used being a poor representation of the overall population (Trochim, Donnelly, & Arora 2016). This is why the sampling method is an identified method to establish external validity.

The sampling method while beneficial is a threat to external validity. With the sampling method, it is difficult to know what section of the population should be used (Trochim, Donnelly, & Arora 2016). Just selecting a sample from the population without knowing what section of the population to use, causes there to be the possibility of a lack of fair representation (Trochim, Donnelly, & Arora 2016). The explanations that identify how the specified generalizations could be wrong can be wrong in terms of the setting, population, time, etc. (Trochim, Donnelly, & Arora 2016).

Construct Validity

Threats to construct validity are reactivity, lack of thought, mono-operation bias, and mono-method bias (Cozby 2014; Trochim, Donnelly, & Arora 2016). Reactivity measures are reactive if the knowledge of being measured is known and makes measuring behavior more difficult (Cozby 2014). Lack of thought given to initial application and explanations of what the constructs were intended to be (Trochim, Donnelly, & Arora 2016). Mono-operation bias is a threat when there is only one treatment used (Trochim, Donnelly, & Arora 2016). Mono-method bias occurs when there is only one measure of a construct (Trochim, Donnelly, & Arora 2016).

Impact on research

 Each of the discussed types of validity can impact the envisioned research in different ways. External validity can impact the selected research on the impact CSR disclosure reporting has on corporate reputation.  External validity can impact the intended research if the generalizations drawn based on the specified population or setting, cannot be replicated in similar settings or populations. It can also impact the research if the generalizations drawn for that population do not identify with the entire population, but just a portion of the population.

 Internal validity can pose a threat to the selected research if the cause and effect relationship that is assumed to exist, does not exist. The intended research assumes that CSR disclosure reporting impacts corporate reputation. Internal validity can impact this because there could be no direct relationship between CSR disclosure reporting and corporate reputation. The measurements used to measure the relationship between the two constructs can also impact the research if the right measurements are not used.

Conclusion

 Validity affects how information is perceived when provided to individuals or groups (St. Onge, Young, Eva, & Hodges 2017). The different forms of validity explained above can strongly impact the research. Since the research is focused more on a cause and effect relationship, internal validity will be stronger than external validity. This will impact the generalization that can be drawn from the findings.

References

• Andrade, C. (2018). Internal, external, and ecological validity in research design, conduct, and evaluation. Indian Journal of Psychological Medicine, 40(5), 498-499.
• Axjonow, A., Ernstberger, J., & Pott, C. (2018). The impact of corporate social responsibility disclosure on corporate reputation: A non-professional stakeholder perspective. Journal of Business Ethics, 151(2), 429-450.
• Cozby, P. C. (2014). Methods in behavioral research (12^th ed.). Boston, MA: McGraw Hill Higher Education.
• Mishra, S., & Modi, S. B. (2016). Corporate social responsibility and shareholder wealth: The role of marketing capability. Journal of Marketing, 80(1), 26-46. Doi: 10.1509/jm.15.0013
• Person, O. (2012). Incorporating corporate social responsibility and sustainability into a business course: A shared experience. Journal of Education for Business, 87(2), 63-72. Doi: 10.1080/08832323.2011.562933
• St-Onge, C., Young, M., Eva, K. W., & Hodges, B. (2017). Validity: One word with a plurality of meanings. Advances in Health Sciences Education, 22(4), 853-867.
• Trochim, W., Donnelly, J., & Arora, K. (2016). Research methods: The essential knowledge base (2^nd ed.). Mason, OH: Cengage.